CN111988613B - Screen content video quality analysis method based on tensor decomposition - Google Patents

Screen content video quality analysis method based on tensor decomposition Download PDF

Info

Publication number
CN111988613B
CN111988613B CN202010778526.8A CN202010778526A CN111988613B CN 111988613 B CN111988613 B CN 111988613B CN 202010778526 A CN202010778526 A CN 202010778526A CN 111988613 B CN111988613 B CN 111988613B
Authority
CN
China
Prior art keywords
screen content
content video
video sequence
principal component
slice
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010778526.8A
Other languages
Chinese (zh)
Other versions
CN111988613A (en
Inventor
曾焕强
黄海靓
陈婧
侯军辉
曹九稳
张云
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huaqiao University
Original Assignee
Huaqiao University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Huaqiao University filed Critical Huaqiao University
Priority to CN202010778526.8A priority Critical patent/CN111988613B/en
Publication of CN111988613A publication Critical patent/CN111988613A/en
Application granted granted Critical
Publication of CN111988613B publication Critical patent/CN111988613B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/154—Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N17/00—Diagnosis, testing or measuring for television systems or their details

Abstract

The invention relates to a screen content video quality analysis method based on tensor decomposition, which comprises the following steps: carrying out tensor decomposition on the selected reference screen content video sequence and the distorted screen content sequence respectively to obtain principal component slices of a three-direction slice set; respectively extracting Gabor characteristic graphs of the three-direction reference principal component slice and the three-direction distortion principal component slice, and calculating to obtain a three-direction characteristic similarity graph; and obtaining a final distorted screen content video quality analysis value based on the three-direction feature similarity graph. The method fully utilizes the tensor decomposition theory to describe the basic texture structure of the screen content video, extracts edge information highly sensitive to human eyes through the Gabor filter, reflects the subjective perception of a human eye vision system on the screen content video, and has better distorted screen content video quality analysis performance.

Description

Screen content video quality analysis method based on tensor decomposition
Technical Field
The invention belongs to the field of video processing, relates to a video quality analysis method, and particularly relates to a screen content video quality analysis method based on tensor decomposition.
Background
With the rapid development of mobile internet and multimedia technology, screen content video attracts extensive attention in academic and industrial fields, and is widely applied to cloud computing, distance education, online live broadcast, video conferencing and other applications. Different from the traditional natural scene video, the screen content video not only contains continuous tone areas obtained by shooting through a camera, such as photos, natural scene video and the like, but also contains non-continuous tone areas obtained based on a computer, such as characters, diagrams, two-dimensional codes and the like, and also contains motion information with rich changes.
As with conventional natural scene video, screen content video inevitably introduces various distortions during generation, processing, compression, storage, transmission, and rendering, resulting in a reduction in visual effect. Since human eyes are the final recipients of screen content videos, it is necessary to provide a quality analysis model that can quickly and accurately reflect the subjective perceptibility of the screen content videos by the human visual system. However, most of the existing quality analysis methods are designed for the traditional natural scene video and are not suitable for the quality analysis of the screen content video. There is currently a lack of methods for efficient quality analysis of screen content video in the field of video processing. Therefore, the method for analyzing the screen content video quality according with the human eye visual characteristics has important theoretical research significance and practical application value.
Disclosure of Invention
The invention aims to break through the limitation of the prior art and provides a screen content video quality analysis method based on tensor decomposition.
The technical scheme adopted by the invention for solving the technical problem is as follows:
the screen content video quality analysis method based on tensor decomposition comprises the following steps:
input reference screen content video sequence VrAnd distorted screen content video sequence Vd；
For reference screen content video sequence VrAnd distorted screen content video sequence VdCarrying out tensor decomposition to obtain three-direction reference principal component slice Mr,x、Mr,y、Mr,tAnd three-directional distortion principal component slice Md,x、Md,y、Md,t；
Respectively extracting three-direction reference principal component slices Mr,x、Mr,y、Mr,tGabor profile Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-dimensional distortion principal component slice Md,x、Md,y、Md,tGabor profile Fd,x(x,y)、Fd,y(x,y)、Fd,t(x,y)；
Calculating three-direction reference Gabor characteristic diagram Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-direction distortion Gabor characteristic diagram Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) feature similarity graph Sx(x,y)、Sy(x,y)、St(x,y)；
Similarity graph S based on three-direction featuresx(x,y)、Sy(x,y)、St(x, y) obtaining a final distorted screen content video quality analysis value.
Preferably, for reference screen content video sequence VrAnd distorted screen content video sequence VdCarrying out tensor decomposition to obtain three-direction reference principal component slice Mr,x、Mr,y、Mr,tAnd three-directional distortion principal component slice Md,x、Md,y、Md,tThe method comprises the following steps:
step 2.1: video sequence V to be referenced to screen contentrIs regarded as a third-order tensor, and is converted into a core tensor through tensor decomposition
Figure BDA0002619355370000022
And three factor matrices Ar,Br,CrThe combination of (a) and (b) is specifically as follows:
Figure BDA0002619355370000023
wherein the extract isnDenotes n-modulo multiplication, n =1,2,3, three factor matrices ar,Br,CrRespectively representing the original video sequence VrPrincipal components in x, y and t directions, which are orthogonal to each other, and core tensor
Figure BDA0002619355370000024
Is represented as follows:
Figure BDA0002619355370000021
video sequence V with distorted screen contentdIs regarded as a third-order tensor, and is converted into a core tensor through tensor decomposition
Figure BDA0002619355370000025
And three factor matrices Ad,Bd,CdThe combination of (a) and (b) is specifically as follows:
Figure BDA0002619355370000026
wherein the extract isnDenotes n-modulo multiplication, n =1,2,3, three factor matrices ad,Bd,CdRespectively representing the original video sequence VdPrincipal components in x, y and t directions, which are orthogonal to each other, and core tensor
Figure BDA0002619355370000027
Is represented as follows:
Figure BDA0002619355370000031
step 2.2: respectively setting a reference factor matrix ArAnd distortion factor matrix AdFor the identity matrix, a reference screen content video sequence V is obtainedrAnd distorted screen content video sequence VdSet of vertical spatiotemporal slices cut along the x-axis direction as follows:
Figure BDA0002619355370000032
Figure BDA0002619355370000033
respectively setting reference factor matrixes BrAnd distortion factor momentArray BdFor the identity matrix, a reference screen content video sequence V is obtainedrAnd distorted screen content video sequence VdSet of horizontal spatiotemporal slices cut along the y-axis direction as follows:
Figure BDA0002619355370000034
Figure BDA0002619355370000035
respectively setting reference factor matrixes CrAnd distortion factor matrix CdFor the identity matrix, a reference screen content video sequence V is obtainedrAnd distorted screen content video sequence VdThe set of spatial slices cut along the t-axis direction is as follows:
Figure BDA0002619355370000036
Figure BDA0002619355370000037
step 2.3: extracting a reference screen content video sequence VrThe slice with the largest energy in the three-direction slice set is used as a reference principal component slice Mr,x、Mr,y、Mr,tThe method comprises the following steps:
Figure BDA0002619355370000038
Figure BDA0002619355370000039
Figure BDA00026193553700000310
extracting a distorted screen content video sequence VdThe slice with the largest energy in the three-direction slice set is used as a reference principal component slice Md,x、Md,y、Md,tThe method comprises the following steps:
Figure BDA00026193553700000311
Figure BDA00026193553700000312
Figure BDA00026193553700000313
wherein w =1,2,...,W，h＝1,2,...,H，l＝1,2, L, W, H, L respectively represent the number of slices of the three-directional slice set.
Preferably, a three-direction reference Gabor characteristic diagram F is calculatedr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-direction distortion Gabor characteristic diagram Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) feature similarity graph Sx(x,y)、Sy(x,y)、St(x, y), as follows:
respectively extracting three-direction reference principal component slices Mr,x、Mr,y、Mr,tGabor profile Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) as follows:
Figure BDA0002619355370000041
Figure BDA0002619355370000042
Figure BDA0002619355370000043
wherein, Gi(x, y) is a Gabor filter as follows:
Figure BDA0002619355370000044
x′＝xcosθ+ysinθ
y′＝y cosθ-xsinθ
where (x, y) denotes coordinates of each pixel in the input principal component slice, i denotes a direction index of the Gabor filter, f and θ are frequency amplitude and direction information of the sinusoidal plane wave (x ', y'), σxAnd σyThe standard deviation of the gaussian kernel in the x-axis direction and the y-axis direction, respectively, is taken here as f =0.2, σx＝2.15,σy=0.15.n is the total number of directions, where a total of 12 directions are considered, corresponding to Gabor filters of θ = i × pi/12, i ∈ { 0.. 11}, respectively.
Respectively extracting three-direction distortion principal component slices Md,x、Md,y、Md,tGabor profile Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) as follows:
Figure BDA0002619355370000045
Figure BDA0002619355370000046
Figure BDA0002619355370000047
preferably, a three-direction reference Gabor characteristic diagram F is calculatedr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-direction distortion Gabor characteristic diagram Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) feature similarity graph Sx(x,y)、Sy(x,y)、St(x, y), as follows:
Figure BDA0002619355370000051
Figure BDA0002619355370000052
Figure BDA0002619355370000053
where c is a constant to ensure numerical stability, c =1000.
Preferably, the similarity graph S is based on three-direction characteristicx(x,y)、Sy(x,y)、St(x, y) obtaining a final distorted screen content video quality analysis value, which is as follows:
by pooling the x-direction feature similarity map Sx(x, y) obtaining an x-direction distortion screen content video quality score:
ωx(x,y)＝max{|Fr,x(x,y)|,|Fd,x(x,y)|}
Figure BDA0002619355370000054
by pooling the y-direction feature similarity map Sy(x, y) deriving a y-direction distortion screen content video quality score:
ωy(x,y)＝max{|Fr,y(x,y)|,|Fd,y(x,y)}
Figure BDA0002619355370000055
by pooling t-direction feature similarity maps Sx(x, y) obtaining a t-direction distortion screen content video quality score:
ωt(x,y)＝max{|Fr,t(x,y)|,|Fd,t(x,y)|}
Figure BDA0002619355370000056
combining the quality scores of the distorted screen contents in all directions to obtain a final distorted screen content video quality analysis value:
Score＝scorex·scorey·scoret
the invention has the following beneficial effects:
the invention provides a screen content video quality analysis method based on tensor decomposition. The method focuses on fully considering the characteristics of a human eye vision system and the characteristics of screen content videos, adopts tensor decomposition to obtain main texture structure information of the screen content videos, fully utilizes Gabor characteristics to capture edge information highly sensitive to human eyes, reflects the subjective perception of the human eye vision subjective vision system on the screen content videos, and has better screen content video quality analysis performance.
Drawings
FIG. 1 is a schematic flow diagram of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Referring to fig. 1, a method for analyzing the video quality of the screen content based on tensor decomposition includes the following specific steps:
step 1, inputting a reference screen content video sequence VrAnd distorted screen content video sequence Vd；
Step 2, for the reference screen content video sequence VrAnd distorted screen content video sequence VdCarrying out tensor decomposition to obtain three-direction reference principal component slice Mr,x、Mr,y、Mr,tAnd three-directional distortion principal component slice Md,x、Md,y、Md,tThe method comprises the following steps:
step 2.1: video sequence V to be referenced to screen contentrIs regarded as a third-order tensor, and is converted by tensor decompositionConversion to a core tensor
Figure BDA0002619355370000063
And three factor matrices Ar,Br,CrThe combination of (a) and (b) is specifically as follows:
Figure BDA0002619355370000064
wherein the extract isnDenotes n-modulo multiplication, n =1,2,3, three factor matrices ar,Br,CrRespectively representing an original video sequence VrPrincipal components in x, y and t directions, which are orthogonal to each other, and core tensor
Figure BDA0002619355370000065
Is represented as follows:
Figure BDA0002619355370000061
video sequence V with distorted screen contentdIs regarded as a third-order tensor, and is converted into a core tensor through tensor decomposition
Figure BDA0002619355370000068
And three factor matrices Ad,Bd,CdThe combination of (a) and (b) is specifically as follows:
Figure BDA0002619355370000066
wherein the extract isnDenotes n-modulo multiplication, n =1,2,3, three factor matrices ad,Bd,CdRespectively representing the original video sequence VdPrincipal components in x, y and t directions, which are orthogonal to each other, and core tensor
Figure BDA0002619355370000067
Is represented as follows:
Figure BDA0002619355370000062
step 2.2: respectively setting a reference factor matrix ArAnd distortion factor matrix AdFor the identity matrix, a reference screen content video sequence V is obtainedrAnd distorted screen content video sequence VdSet of vertical spatiotemporal slices cut along the x-axis direction as follows:
Figure BDA0002619355370000071
Figure BDA0002619355370000072
respectively setting reference factor matrixes BrAnd distortion factor matrix BdFor the identity matrix, a reference screen content video sequence V is obtainedrAnd distorted screen content video sequence VdSet of horizontal spatiotemporal slices cut along the y-axis direction as follows:
Figure BDA0002619355370000073
Figure BDA0002619355370000074
respectively setting reference factor matrixes CrAnd distortion factor matrix CdFor the identity matrix, a reference screen content video sequence V is obtainedrAnd distorted screen content video sequence VdThe set of spatial slices cut along the t-axis direction is as follows:
Figure BDA0002619355370000075
Figure BDA0002619355370000076
step 2.3: extracting a reference screen content video sequence VrThe slice with the largest energy in the three-direction slice set is used as a reference principal component slice Mr,x、Mr,y、Mr,tThe method comprises the following steps:
Figure BDA0002619355370000077
Figure BDA0002619355370000078
Figure BDA0002619355370000079
extraction of distorted screen content video sequence VdThe slice with the largest energy in the three-direction slice set is used as a reference principal component slice Md,x、Md,y、Md,tThe method comprises the following steps:
Figure BDA00026193553700000710
Figure BDA00026193553700000711
Figure BDA00026193553700000712
wherein w =1,2,...,W，h＝1,2,...,H，l＝1,2, L, W, H, L respectively represent the number of slices of the three-directional slice set.
Step 3, calculating a three-direction reference Gabor characteristic diagram Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-direction distortion Gabor feature map Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) feature similarity graph Sx(x,y)、Sy(x,y)、St(x, y), as follows:
respectively extracting three-direction reference principal component slices Mr,x、Mr,y、Mr,tGabor profile Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) as follows:
Figure BDA0002619355370000081
Figure BDA0002619355370000082
Figure BDA0002619355370000083
wherein G isi(x, y) is a Gabor filter as follows:
Figure BDA0002619355370000084
x′＝xcosθ+ysinθ
y′＝y cosθ-xsinθ
where (x, y) denotes coordinates of each pixel in the input principal component slice, i denotes a direction index of the Gabor filter, f and θ are frequency amplitude and direction information of the sinusoidal plane wave (x ', y'), σxAnd σyThe standard deviation of the gaussian kernel in the x-axis direction and the y-axis direction, respectively, is taken here as f =0.2, σx＝2.15,σy=0.15.n is the total number of directions, where a total of 12 directions are considered, corresponding to Gabor filters of θ = i × pi/12, i ∈ { 0.. 11}, respectively.
Respectively extracting three-direction distortion principal component slices Md,x、Md,y、Md,tGabor profile Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) as follows:
Figure BDA0002619355370000085
Figure BDA0002619355370000086
Figure BDA0002619355370000087
step 4, calculating a three-direction reference Gabor characteristic diagram Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-direction distortion Gabor characteristic diagram Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) feature similarity graph Sx(x,y)、Sy(x,y)、St(x, y), as follows:
Figure BDA0002619355370000091
Figure BDA0002619355370000092
Figure BDA0002619355370000093
where c is a constant to ensure numerical stability, c =1000.
Step 5, similarity graph S based on three-direction characteristicsx(x,y)、Sy(x,y)、St(x, y) obtaining a final distorted screen content video quality analysis value, which is as follows:
by pooling the x-direction feature similarity map Sx(x, y) obtaining an x-direction distortion screen content video quality score:
ωx(x,y)＝max{|Fr,x(x,y)|,|Fd,x(x,y)|}
Figure BDA0002619355370000094
by pooling the y-direction feature similarity map Sy(x, y) deriving a y-direction distortion screen content video quality score:
ωy(x,y)＝max{|Fr,y(x,y)|,|Fd,y(x,y)}
Figure BDA0002619355370000095
by pooling t-direction feature similarity map Sx(x, y) obtaining a t-direction distortion screen content video quality score:
ωt(x,y)＝max{|Fr,t(x,y)|,|Fd,t(x,y)|}
Figure BDA0002619355370000096
combining the quality scores of the distorted screen contents in all directions to obtain a final distorted screen content video quality analysis value:
Score＝scorex·scorey·scoret
the above examples are provided only for illustrating the present invention and are not intended to limit the present invention. Changes, modifications, etc. to the above-described embodiments are intended to fall within the scope of the claims of the present invention as long as they are in accordance with the technical spirit of the present invention.

Claims (4)

1. A screen content video quality analysis method based on tensor decomposition is characterized by comprising the following steps:
inputting a reference screen content video sequence VrAnd distorted screen content video sequence Vd；
For reference screen content video sequence VrAnd distorted screen content video sequence VdCarrying out tensor decomposition to obtain three-direction reference principal component slice Mr,x、Mr,y、Mr,tAnd three-directional distortion principal component slice Md,x、Md,y、Md,t；
Respectively extracting three-direction reference principal component slices Mr,x、Mr,y、Mr,tGabor profile Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-dimensional distortion principal component slice Md,x、Md,y、Md,tGabor profile Fd,x(x,y)、Fd,y(x,y)、Fd,t(x,y)；
Calculating three-direction reference Gabor characteristic diagram Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-direction distortion Gabor characteristic diagram Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) feature similarity graph Sx(x,y)、Sy(x,y)、St(x,y)；
Similarity graph S based on three-direction featuresx(x,y)、Sy(x,y)、St(x, y) obtaining a final distorted screen content video quality analysis value;
for reference screen content video sequence VrAnd distorted screen content video sequence VdCarrying out tensor decomposition to obtain three-direction reference principal component slice Mr,x、Mr,y、Mr,tAnd three-directional distortion principal component slice Md,x、Md,y、Md,tThe method comprises the following steps:
step 2.1: video sequence V to be referenced to screen contentrIs regarded as a third-order tensor, and is converted into a core tensor through tensor decomposition
Figure FDA0003834220680000012
And three factor matrices Ar,Br,CrThe combination of (a) and (b) is specifically as follows:
Figure FDA0003834220680000013
wherein the extract isnDenotes n-modulo multiplication, n =1,2,3, three factor matrices ar,Br,CrRespectively representing the original video sequence VrPrincipal components in x, y and t directions, which are orthogonal to each other, and core tensor
Figure FDA0003834220680000014
Is represented as follows:
Figure FDA0003834220680000011
video sequence V with distorted screen contentdIs regarded as a third-order tensor, and is converted into a core tensor through tensor decomposition
Figure FDA0003834220680000015
And three factor matrices Ad,Bd,CdThe combination of (a) and (b) is specifically as follows:
Figure FDA00038342206800000211
wherein the extract isnDenotes n-modulo multiplication, n =1,2,3, three factor matrices ad,Bd,CdRespectively representing the original video sequence VdPrincipal components in x, y and t directions, which are orthogonal to each other, and core tensor
Figure FDA00038342206800000212
Is represented as follows:
Figure FDA0003834220680000021
step 2.2: respectively setting a reference factor matrix ArAnd distortion factor matrix AdFor the identity matrix, a reference screen is obtainedContent video sequence VrAnd distorted screen content video sequence VdSet of vertical spatiotemporal slices cut along the x-axis direction as follows:
Figure FDA0003834220680000022
Figure FDA0003834220680000023
respectively setting reference factor matrixes BrAnd distortion factor matrix BdFor the identity matrix, a reference screen content video sequence V is obtainedrAnd distorted screen content video sequence VdSet of horizontal spatiotemporal slices cut along the y-axis direction as follows:
Figure FDA0003834220680000024
Figure FDA0003834220680000025
respectively setting reference factor matrixes CrAnd distortion factor matrix CdFor the identity matrix, a reference screen content video sequence V is obtainedrAnd distorted screen content video sequence VdThe set of spatial slices cut along the t-axis direction is as follows:
Figure FDA0003834220680000026
Figure FDA0003834220680000027
step 2.3: extracting a reference screen content video sequence VrThree-directional slice collection ofThe slice with the highest energy is used as the reference principal component slice Mr,x、Mr,y、Mr,tThe method comprises the following steps:
Figure FDA0003834220680000028
Figure FDA0003834220680000029
Figure FDA00038342206800000210
extraction of distorted screen content video sequence VdThe slice with the largest energy in the three-direction slice set is used as a reference principal component slice Md,x、Md,y、Md,tThe method comprises the following steps:
Figure FDA0003834220680000031
Figure FDA0003834220680000032
Figure FDA0003834220680000033
where W =1,2., W, H =1,2., H, L =1,2., L, W, H, L respectively represent the number of slices of a three-directional slice set.
2. The tensor decomposition-based screen content video quality analysis method as recited in claim 1, wherein three-directional reference principal component slices M are respectively extractedr,x、Mr,y、Mr,tGabor profile Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-dimensional distortion principal component slice Md,x、Md,y、Md,tGabor feature map of (1)d,x(x,y)、Fd,y(x,y)、Fd,t(x, y), as follows:
respectively extracting three-direction reference principal component slices Mr,x、Mr,y、Mr,tGabor profile Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) as follows:
Figure FDA0003834220680000034
Figure FDA0003834220680000035
Figure FDA0003834220680000036
wherein G isi(x, y) is a Gabor filter as follows:
Figure FDA0003834220680000037
x′＝xcosθ+ysinθ
y′＝ycosθ-xsinθ
where (x, y) denotes coordinates of each pixel in the input principal component slice, i denotes a direction index of the Gabor filter, f and θ are frequency amplitude and direction information of the sinusoidal plane wave (x ', y'), σxAnd σyThe standard deviation of the gaussian kernel in the x-axis direction and the y-axis direction, respectively, is taken here as f =0.2, σx＝2.15,σy=0.15; n is the total number of directions, and a total of 12 directions are considered here, which respectively correspond to Gabor filters of theta = i × π/12, i ∈ { 0.. 11 };
respectively extracting three-direction distortion principal component slices Md,x、Md,y、Md,tGabor profile Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) as follows:
Figure FDA0003834220680000041
Figure FDA0003834220680000042
Figure FDA0003834220680000043
3. the tensor decomposition-based screen content video quality analysis method as recited in claim 1, wherein: calculating three-direction reference Gabor characteristic diagram Fr,x(x,y)、Fr,y(x,y)、Fr,t(x, y) and three-direction distortion Gabor characteristic diagram Fd,x(x,y)、Fd,y(x,y)、Fd,t(x, y) feature similarity graph Sx(x,y)、Sy(x,y)、St(x, y), as follows:
Figure FDA0003834220680000044
Figure FDA0003834220680000045
Figure FDA0003834220680000046
where c is a constant to ensure numerical stability, c =1000.
4. The tensor decomposition-based screen content video quality analysis method as recited in claim 1, wherein: similarity graph S based on three-direction featuresx(x,y)、Sy(x,y)、St(x, y) obtaining a final distorted screen content video quality analysis value, which is as follows:
by pooling the x-direction feature similarity map Sx(x, y) obtaining an x-direction distortion screen content video quality score:
ωx(x,y)＝max{Fr,x(x,y),Fd,x(x,y)}
Figure FDA0003834220680000047
by pooling the y-direction feature similarity map Sy(x, y) deriving a y-direction distortion screen content video quality score:
ωy(x,y)＝max{Fr,y(x,y),Fd,y(x,y)}
Figure FDA0003834220680000051
by pooling t-direction feature similarity maps Sx(x, y) obtaining a t-direction distortion screen content video quality score:
ωt(x,y)＝max{Fr,t(x,y),Fd,t(x,y)}
Figure FDA0003834220680000052
and combining the quality scores of the distorted screen contents in all directions to obtain a final distorted screen content video quality analysis value:
Score＝scorex·scorey·scoret。
CN202010778526.8A 2020-08-05 2020-08-05 Screen content video quality analysis method based on tensor decomposition Active CN111988613B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010778526.8A CN111988613B (en) 2020-08-05 2020-08-05 Screen content video quality analysis method based on tensor decomposition

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010778526.8A CN111988613B (en) 2020-08-05 2020-08-05 Screen content video quality analysis method based on tensor decomposition

Publications (2)

Publication Number Publication Date
CN111988613A CN111988613A (en) 2020-11-24
CN111988613B true CN111988613B (en) 2022-11-01

Family

ID=73446030

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010778526.8A Active CN111988613B (en) 2020-08-05 2020-08-05 Screen content video quality analysis method based on tensor decomposition

Country Status (1)

Country Link
CN (1) CN111988613B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012000136A1 (en) * 2010-07-02 2012-01-05 Thomson Broadband R&D (Beijing) Co., Ltd. Method for measuring video quality using a reference, and apparatus for measuring video quality using a reference
CN102737380A (en) * 2012-06-05 2012-10-17 宁波大学 Stereo image quality objective evaluation method based on gradient structure tensor
CN110958449A (en) * 2019-12-19 2020-04-03 西北工业大学 Three-dimensional video subjective perception quality prediction method

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9794554B1 (en) * 2016-03-31 2017-10-17 Centre National de la Recherche Scientifique—CNRS Method for determining a visual quality index of a high dynamic range video sequence

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2012000136A1 (en) * 2010-07-02 2012-01-05 Thomson Broadband R&D (Beijing) Co., Ltd. Method for measuring video quality using a reference, and apparatus for measuring video quality using a reference
CN102737380A (en) * 2012-06-05 2012-10-17 宁波大学 Stereo image quality objective evaluation method based on gradient structure tensor
CN110958449A (en) * 2019-12-19 2020-04-03 西北工业大学 Three-dimensional video subjective perception quality prediction method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Novel Spatio-Temporal Structural Information Based Video Quality Metric;Yue Wang等;《IEEE Transactions on Circuits and Systems for Video Technology》;20120203;全文 *
基于时域梯度相似度的视频质量评价模型;邱亮等;《计算机工程与科学》;20180415(第04期);全文 *

Also Published As

Publication number Publication date
CN111988613A (en) 2020-11-24

Similar Documents

Publication Publication Date Title
US11830230B2 (en) Living body detection method based on facial recognition, and electronic device and storage medium
Hu et al. Learning supervised scoring ensemble for emotion recognition in the wild
CN107844795B (en) Convolutional neural networks feature extracting method based on principal component analysis
CN108898145A (en) A kind of image well-marked target detection method of combination deep learning
CN110827312B (en) Learning method based on cooperative visual attention neural network
CN110136144B (en) Image segmentation method and device and terminal equipment
CN105095857B (en) Human face data Enhancement Method based on key point perturbation technique
CN111091075A (en) Face recognition method and device, electronic equipment and storage medium
CN111209811A (en) Method and system for detecting eyeball attention position in real time
CN113688839B (en) Video processing method and device, electronic equipment and computer readable storage medium
CN111680577A (en) Face detection method and device
CN106295514A (en) A kind of method and device of image recognition exercise question display answer
CN106203448A (en) A kind of scene classification method based on Nonlinear Scale Space Theory
CN113888501B (en) Attention positioning network-based reference-free image quality evaluation method
CN111510707B (en) Full-reference screen video quality evaluation method based on space-time Gabor feature tensor
CN110472567A (en) A kind of face identification method and system suitable under non-cooperation scene
CN111988613B (en) Screen content video quality analysis method based on tensor decomposition
CN105678208B (en) Extract the method and device of face texture
CN110070626B (en) Three-dimensional object retrieval method based on multi-view classification
CN111652238B (en) Multi-model integration method and system
Yang et al. Research on human motion recognition based on data redundancy technology
CN111881794B (en) Video behavior recognition method and system
CN106776838A (en) A kind of massive video analysis and quick retrieval system based on cloud computing
CN113014916B (en) Screen video quality identification method based on local video activity
Agarwal et al. Content based image retrieval based on log Gabor wavelet transform

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant